How to Select and Calculate a Heatsink for an IC, MOSFET, or Other PCB Component

 

 

A heatsink must provide low enough thermal resistance to keep the junction temperature below a safe limit. Wattage or physical size alone won't tell you what to buy. You need to know how much heat is generated, how hot the part may get, and how easily heat travels from the silicon to the air.

1. The heat-flow path

Treat heat flow like current through series resistors, with temperature difference as the voltage. Heat passes through:

1.       Junction (the die), at T_J.

2.       Case, crossed with R_θJC (from the datasheet).

3.       Thermal interface (grease, pad, insulator), R_θCS.

4.       Heatsink, R_θSA.

5.       Ambient air, at T_A.

Resistances are in °C/W (K/W). A poor interface can ruin an otherwise good heatsink.

Limits of the series model: it assumes essentially all heat leaves through the case, interface, and heatsink. Many packages also shed heat through other paths in parallel, such as leads and PCB copper (a TO-220's leads, or an SMD part cooled from both the board and a top-mounted sink). Ignoring these makes the estimate conservative, but for such devices the simple series model may not capture the full thermal behavior. Use the manufacturer's thermal model where one is given.

2. Step 1: Find the power dissipated as heat

Use the power lost inside the component, not the power delivered to the load.

MOSFET conduction loss:

P_cond = I_D² × R_DS(on)

Use R_DS(on) at your actual gate-drive voltage and operating junction temperature, not the 25 °C headline value. Check the datasheet's resistance-versus-temperature curve. Hot values are often 1.5 to 2 times higher. At 5 A with 0.04 Ω at 25 °C, the loss is 1.0 W. If the hot value is 0.064 Ω, it is 1.6 W, which is 60% higher than the cold estimate. Designing around the cold figure significantly underestimates the heat.

MOSFET switching loss (first-order estimate):

P_sw ≈ ½ × V_DS × I_D × (t_r + t_f) × f_sw

This assumes idealized linear transitions. Real losses depend on the actual voltage and current waveforms, gate resistance, parasitic capacitances, reverse recovery, and load type (inductive loads usually cost more). Body-diode conduction and reverse-recovery losses may also apply.

Output-capacitance loss:

P_oss ≈ E_oss × f_sw

C_oss varies strongly with drain voltage, so the simple ½ × C_oss × V² × f_sw form is only a rough guide. Where the datasheet gives the stored energy E_oss (at or near your operating voltage), use it instead.

Gate-drive power:

P_g ≈ Q_g × V_GS × f_sw

This is the power needed to charge and discharge the gate each second. It is dissipated across the driver's output stage and the gate resistor, not only in the driver. It matters for driver heating but is usually small for the MOSFET's own heatsink. At high frequency, rely on the manufacturer's loss tools or measured waveforms.

Linear regulator or linear IC:

P_D ≈ (V_IN − V_OUT) × I_OUT + V_IN × I_Q

Example: 12 V to 5 V at 0.2 A with 5 mA quiescent current gives 1.4 + 0.06 = 1.46 W. This form is valid when I_Q is the additional ground/quiescent current drawn from the input. Datasheets don't all define current the same way, so verify the definition. For a regulator with significant ground-pin current, or one whose quiescent current varies with load, use the actual input and output currents:

P_D = V_IN × I_IN − V_OUT × I_OUT

Other parts: resistors dissipate I²R or V²/R. BJTs dissipate roughly V_CE × I_C plus base-drive and switching losses. For a switching converter, use the switching device's actual losses, not the load power.

3. Step 2: Find the maximum allowable heatsink resistance

T_J = T_A + P_D × (R_θJC + R_θCS + R_θSA)

Solving for the heatsink:

R_θSA ≤ (T_J,target − T_A) / P_D − R_θJC − R_θCS

Choosing the target: don't design to the absolute maximum (often 150 or 175 °C). Typical derated targets are 100 to 125 °C for commercial designs and 80 to 100 °C for high-reliability or hot-environment equipment. Use the worst-case air temperature near the part, not room temperature.

Typical R_θCS (approximate; depends on package, contact area, interface thickness, and clamping pressure):

•        Grease on a flat, clamped surface: about 0.1 to 0.5 °C/W.

•        Silicone pad: about 0.5 to 1.5 °C/W.

•        Mica insulator plus grease: about 1 to 3 °C/W or more.

Worked example: MOSFET at 3 W

Assume 3 W (hot R_DS(on) included), 40 °C worst-case ambient, 150 °C absolute maximum, 110 °C design target, R_θJC = 2 °C/W, and R_θCS = 0.5 °C/W.

At the absolute maximum:

R_θSA ≤ (150 − 40) / 3 − 2 − 0.5 = 36.67 − 2.5 = 34.17 °C/W

At the design target, which is the number to use:

R_θSA ≤ (110 − 40) / 3 − 2 − 0.5 = 23.33 − 2.5 = 20.83 °C/W

Real-world conditions. Published heatsink ratings are measured under specific conditions (airflow, orientation, temperature rise), and performance in an enclosure is often worse. Use the manufacturer's thermal curves for your actual conditions where available. Otherwise apply an engineering margin and verify by testing. For illustration only, assume a 25% penalty (an assumption, not a universal factor):

20.83 / 1.25 ≈ 16.7 °C/W

A 15 °C/W heatsink passes this check. With the assumed penalty its effective resistance is 18.75 °C/W:

T_J = 40 + 3 × (2 + 0.5 + 18.75) = 103.75 °C

At exactly 15 °C/W, T_J = 92.5 °C. Either way it is under the 110 °C target.

Practical check. Under the same assumptions, the predicted case temperature is:

T_C = T_A + P_D × (R_θCS + R_θSA) = 40 + 3 × (0.5 + 18.75) = 97.75 °C

This is valid only for the assumed steady-state model and measurement location. When comparing against a measurement:

•        Measure at the defined case reference point (usually where the datasheet specifies R_θJC, such as the tab or exposed pad), with good thermal contact for the probe.

•        Wait for thermal equilibrium before reading.

•        Remember the junction is hotter than the case by P_D × R_θJC, so a measured case temperature must be converted using the actual power and thermal path.

A measured value well above prediction means the real dissipation or the mounting is worse than assumed.

4. Components mounted directly on a PCB

Through-hole (TO-220, TO-247): attach a heatsink to the tab. Check whether the tab is a live terminal (drain, collector); if so, an insulator is needed, which raises R_θCS.

SMD (SO-8, DPAK, PowerPAK, QFN): heat flows mainly into PCB copper. Three datasheet parameters matter:

•        R_θJA: junction-to-ambient under a specified test board and setup. It is board-dependent and valid only for similar boards.

•        R_θJC: junction-to-case (or to the exposed pad).

•        R_θJB: junction-to-board, where specified.

Adding an external heatsink does not make the datasheet R_θJA apply; use R_θJC plus the external path instead. For PCB-cooled parts, follow the manufacturer's recommended pad layout and thermal guidance. Cooling is improved by:

•        More copper connected to the thermal pad or power terminal.

•        Thermal vias under the exposed pad (effectiveness depends on pad geometry, via size and count, and layer stack-up).

•        Thicker copper where practical.

•        Airflow and spacing from other hot parts.

5. Choosing the physical heatsink

Compare candidates by published R_θSA under conditions like yours. Rough natural-convection guides: small TO-220 clip-ons are about 20 to 30 °C/W or worse, medium extrusions about 5 to 10 °C/W, and large finned sinks a few °C/W. These are illustrations only. Check:

1.       Natural vs. forced convection (a fan-rated figure needs the fan).

2.       Fin orientation (vertical fins suit natural convection).

3.       Worst-case local ambient inside the enclosure.

4.       Mounting interface: flatness, compound, and clamping pressure.

5.       Electrical isolation from live terminals.

6.       Multiple devices on one sink: add their powers for R_θSA, then check each junction separately with its own R_θJC and interface.

7.       Neighboring heat sources.

A heatsink's wattage rating alone isn't enough. The relevant spec is thermal resistance under conditions representative of your installation.

6. Pulsed and transient loads

A steady-state calculation alone does not establish that a pulsed condition is safe. For pulses shorter than the time needed to reach thermal equilibrium, the junction rise can be lower than the steady-state prediction. Use the datasheet's transient thermal impedance Z_θJC(t) to estimate peak junction temperature. For repetitive pulses, account for pulse width, duty cycle, and residual heat from preceding pulses; at high duty cycles the result approaches steady state. Also check the safe operating area (SOA) and pulsed-power limits.

7. A warning about datasheet ratings

A part rated for 50 W maximum dissipation cannot necessarily dissipate 50 W on your board. That figure usually assumes the case is held at 25 °C by an ideal heatsink. Your real limit depends on your ambient, thermal path, and junction target.

Quick recipe

1.       Find P_D at hot operating conditions, including switching losses.

2.       Choose worst-case T_A and a derated T_J target.

3.       Look up R_θJC; estimate R_θCS from your mounting method.

4.       Compute R_θSA,max = (T_J,target − T_A) / P_D − R_θJC − R_θCS.

5.       Account for real-world conditions using manufacturer thermal data where available; otherwise, apply an appropriate engineering margin and verify by testing.

6.       Measure case temperature at a defined point under worst-case operation, after reaching thermal equilibrium, and compare it with your prediction.